Re: [PATCH] compiler-gcc.h: Define __SANITIZE_ADDRESS__ under hwaddress sanitizer

[Marco Elver <elver@google.com>]

On Wed, 20 Oct 2021 at 22:00, Kees Cook <keescook@chromium.org> wrote:
> When Clang is using the hwaddress sanitizer, it sets __SANITIZE_ADDRESS__
> explicitly:
>
>  #if __has_feature(address_sanitizer) || __has_feature(hwaddress_sanitizer)
>  /* Emulate GCC's __SANITIZE_ADDRESS__ flag */
>  #define __SANITIZE_ADDRESS__
>  #endif

Hmm, the comment is a little inaccurate if hwaddress sanitizer is on,
but I certainly wouldn't want compiler-clang.h to start emulating gcc
here and start defining __SANITIZE_HWADDRESS__ if the places where we
check it are the same as __SANITIZE_ADDRESS__. So this patch is the
right approach.

> Once hwaddress sanitizer was added to GCC, however, a separate define
> was created, __SANITIZE_HWADDRESS__. The kernel is expecting to find
> __SANITIZE_ADDRESS__ in either case, though, and the existing string
> macros break on supported architectures:
>
>  #if (defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)) && \
>           !defined(__SANITIZE_ADDRESS__)
>
> where as other architectures (like arm32) have no idea about hwaddress
> sanitizer and just check for __SANITIZE_ADDRESS__:
>
>  #if defined(CONFIG_KASAN) && !defined(__SANITIZE_ADDRESS__)

arm32 doesn't support KASAN_SW_TAGS, so I think the bit about arm32 is
irrelevant.

Only arm64 can, and the reason that arm64 doesn't check against
"defined(CONFIG_KASAN)" is because we also have KASAN_HW_TAGS (no
compiler instrumentation).

> This would lead to compiler foritfy self-test warnings when building
> with CONFIG_KASAN_SW_TAGS=y:
>
> warning: unsafe memmove() usage lacked '__read_overflow2' symbol in lib/test_fortify/read_overflow2-memmove.c
> warning: unsafe memcpy() usage lacked '__write_overflow' symbol in lib/test_fortify/write_overflow-memcpy.c
> ...
>
> Sort this out by also defining __SANITIZE_ADDRESS__ in GCC under the
> hwaddress sanitizer.
>
> Suggested-by: Arnd Bergmann <arnd@arndb.de>
> Cc: Nathan Chancellor <nathan@kernel.org>
> Cc: Nick Desaulniers <ndesaulniers@google.com>
> Cc: Miguel Ojeda <ojeda@kernel.org>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Marco Elver <elver@google.com>
> Cc: Will Deacon <will@kernel.org>
> Cc: Arvind Sankar <nivedita@alum.mit.edu>
> Cc: Masahiro Yamada <masahiroy@kernel.org>
> Cc: llvm@lists.linux.dev
> Signed-off-by: Kees Cook <keescook@chromium.org>

Other than that,

  Reviewed-by: Marco Elver <elver@google.com>

Thanks!

> ---
> I'm intending to take this via my overflow series, since that is what introduces
> the compile-test regression tests (which found this legitimate bug). :)
>
> -Kees
> ---
>  include/linux/compiler-gcc.h | 8 ++++++++
>  1 file changed, 8 insertions(+)
>
> diff --git a/include/linux/compiler-gcc.h b/include/linux/compiler-gcc.h
> index 6f24eb8c5dda..ccbbd31b3aae 100644
> --- a/include/linux/compiler-gcc.h
> +++ b/include/linux/compiler-gcc.h
> @@ -121,6 +121,14 @@
>  #define __no_sanitize_coverage
>  #endif
>
> +/*
> + * Treat __SANITIZE_HWADDRESS__ the same as __SANITIZE_ADDRESS__ in the kernel,
> + * matching the defines used by Clang.
> + */
> +#ifdef __SANITIZE_HWADDRESS__
> +#define __SANITIZE_ADDRESS__
> +#endif
> +
>  /*
>   * Turn individual warnings and errors on and off locally, depending
>   * on version.
> --
> 2.30.2
>